Households across the world contain arbitrary objects: from mate gourds and coffee mugs to sitars and guitars. Considering this diversity, robot perception must handle a large variety of semantic objects without additional fine-tuning to be broadly applicable in homes. Recently, zero-shot models have demonstrated impressive performance in image classification of arbitrary objects (i.e., classifying images at inference with categories not explicitly seen during training). In this paper, we translate the success of zero-shot vision models (e.g., CLIP) to the popular embodied AI task of object navigation. In our setting, an agent must find an arbitrary goal object, specified via text, in unseen environments coming from different datasets. Our key insight is to modularize the task into zero-shot object localization and exploration. Employing this philosophy, we design CLIP on Wheels (CoW) baselines for the task and evaluate each zero-shot model in both Habitat and RoboTHOR simulators. We find that a straightforward CoW, with CLIP-based object localization plus classical exploration, and no additional training, often outperforms learnable approaches in terms of success, efficiency, and robustness to dataset distribution shift. This CoW achieves 6.3% SPL in Habitat and 10.0% SPL in RoboTHOR, when tested zero-shot on all categories. On a subset of four RoboTHOR categories considered in prior work, the same CoW shows a 16.1 percentage point improvement in Success over the learnable state-of-the-art baseline.
Parallelizing Gated Recurrent Unit (GRU) networks is a challenging task, as the training procedure of GRU is inherently sequential. Prior efforts to parallelize GRU have largely focused on conventional parallelization strategies such as data-parallel and model-parallel training algorithms. However, when the given sequences are very long, existing approaches are still inevitably performance limited in terms of training time. In this paper, we present a novel parallel training scheme (called parallel-in-time) for GRU based on a multigrid reduction in time (MGRIT) solver. MGRIT partitions a sequence into multiple shorter sub-sequences and trains the sub-sequences on different processors in parallel. The key to achieving speedup is a hierarchical correction of the hidden state to accelerate end-to-end communication in both the forward and backward propagation phases of gradient descent. Experimental results on the HMDB51 dataset, where each video is an image sequence, demonstrate that the new parallel training scheme achieves up to 6.5$\times$ speedup over a serial approach. As efficiency of our new parallelization strategy is associated with the sequence length, our parallel GRU algorithm achieves significant performance improvement as the sequence length increases.
We introduce the problem of predicting, from a single video frame, a low-dimensional subspace of optical flow which includes the actual instantaneous optical flow. We show how several natural scene assumptions allow us to identify an appropriate flow subspace via a set of basis flow fields parameterized by disparity and a representation of object instances. The flow subspace, together with a novel loss function, can be used for the tasks of predicting monocular depth or predicting depth plus an object instance embedding. This provides a new approach to learning these tasks in an unsupervised fashion using monocular input video without requiring camera intrinsics or poses.
The generative adversarial network (GAN) has shown its outstanding capability in improving Non-Autoregressive TTS (NAR-TTS) by adversarially training it with an extra model that discriminates between the real and the generated speech. To maximize the benefits of GAN, it is crucial to find a powerful discriminator that can capture rich distinguishable information. In this paper, we propose a multi-scale time-frequency spectrogram discriminator to help NAR-TTS generate high-fidelity Mel-spectrograms. It treats the spectrogram as a 2D image to exploit the correlation among different components in the time-frequency domain. And a U-Net-based model structure is employed to discriminate at different scales to capture both coarse-grained and fine-grained information. We conduct subjective tests to evaluate the proposed approach. Both multi-scale and time-frequency discriminating bring significant improvement in the naturalness and fidelity. When combining the neural vocoder, it is shown more effective and concise than fine-tuning the vocoder. Finally, we visualize the discriminating maps to compare their difference to verify the effectiveness of multi-scale discriminating.
Recently, with the application of deep learning in the remote sensing image (RSI) field, the classification accuracy of the RSI has been greatly improved compared with traditional technology. However, even state-of-the-art object recognition convolutional neural networks are fooled by the universal adversarial perturbation (UAP). To verify that UAP makes the RSI classification model error classification, this paper proposes a novel method combining an encoder-decoder network with an attention mechanism. Firstly, the former can learn the distribution of perturbations better, then the latter is used to find the main regions concerned by the RSI classification model. Finally, the generated regions are used to fine-tune the perturbations making the model misclassified with fewer perturbations. The experimental results show that the UAP can make the RSI misclassify, and the attack success rate (ASR) of our proposed method on the RSI data set is as high as 97.35%.
Recently, self-supervised methods show remarkable achievements in image-level representation learning. Nevertheless, their image-level self-supervisions lead the learned representation to sub-optimal for dense prediction tasks, such as object detection, instance segmentation, etc. To tackle this issue, several recent self-supervised learning methods have extended image-level single embedding to pixel-level dense embeddings. Unlike image-level representation learning, due to the spatial deformation of augmentation, it is difficult to sample pixel-level positive pairs. Previous studies have sampled pixel-level positive pairs using the winner-takes-all among similarity or thresholding warped distance between dense embeddings. However, these naive methods can be struggled by background clutter and outliers problems. In this paper, we introduce Hough Contrastive Learning (HoughCL), a Hough space based method that enforces geometric consistency between two dense features. HoughCL achieves robustness against background clutter and outliers. Furthermore, compared to baseline, our dense positive pairing method has no additional learnable parameters and has a small extra computation cost. Compared to previous works, our method shows better or comparable performance on dense prediction fine-tuning tasks.
Visual dialog has witnessed great progress after introducing various vision-oriented goals into the conversation, especially such as GuessWhich and GuessWhat, where the only image is visible by either and both of the questioner and the answerer, respectively. Researchers explore more on visual dialog tasks in such kind of single- or perfectly co-observable visual scene, while somewhat neglect the exploration on tasks of non perfectly co-observable visual scene, where the images accessed by two agents may not be exactly the same, often occurred in practice. Although building common ground in non-perfectly co-observable visual scene through conversation is significant for advanced dialog agents, the lack of such dialog task and corresponding large-scale dataset makes it impossible to carry out in-depth research. To break this limitation, we propose an object-referring game in non-perfectly co-observable visual scene, where the goal is to spot the difference between the similar visual scenes through conversing in natural language. The task addresses challenges of the dialog strategy in non-perfectly co-observable visual scene and the ability of categorizing objects. Correspondingly, we construct a large-scale multimodal dataset, named SpotDiff, which contains 87k Virtual Reality images and 97k dialogs generated by self-play. Finally, we give benchmark models for this task, and conduct extensive experiments to evaluate its performance as well as analyze its main challenges.
With the huge expansion of internet and trillions of gigabytes of data generated every single day, the needs for the development of various tools has become mandatory in order to maintain system adaptability to rapid changes. One of these tools is known as Image Captioning. Every entity in internet must be properly identified and managed and therefore in the case of image data, automatic captioning for identification is required. Similarly, content generation for missing labels, image classification and artificial languages all requires the process of Image Captioning. This paper discusses an efficient and unique way to perform automatic image captioning on individual image and discusses strategies to improve its performances and functionalities.
Image Quality Assessment (IQA) is of great value in the workflow of Magnetic Resonance Imaging (MRI)-based analysis. Blind IQA (BIQA) methods are especially required since high-quality reference MRI images are usually not available. Recently, many efforts have been devoted to developing deep learning-based BIQA approaches. However, the performance of these methods is limited due to the utilization of simple content-non-adaptive network parameters and the waste of the important 3D spatial information of the medical images. To address these issues, we design a 3D content-adaptive hyper-network for MRI BIQA. The overall 3D configuration enables the exploration of comprehensive 3D spatial information from MRI images, while the developed content-adaptive hyper-network contributes to the self-adaptive capacity of network parameters and thus, facilitates better BIQA performance. The effectiveness of the proposed method is extensively evaluated on the open dataset, MRIQC. Promising performance is achieved compared with the corresponding baseline and 4 state-of-the-art BIQA methods. We make our code available at \url{https://git.openi.org.cn/SIAT_Wangshanshan/HyS-Net}.
The interpretability of medical image analysis models is considered a key research field. We use a dataset of eye-tracking data from five radiologists to compare the outputs of interpretability methods against the heatmaps representing where radiologists looked. We conduct a class-independent analysis of the saliency maps generated by two methods selected from the literature: Grad-CAM and attention maps from an attention-gated model. For the comparison, we use shuffled metrics, which avoid biases from fixation locations. We achieve scores comparable to an interobserver baseline in one shuffled metric, highlighting the potential of saliency maps from Grad-CAM to mimic a radiologist's attention over an image. We also divide the dataset into subsets to evaluate in which cases similarities are higher.